结合小波系数的normalized cut分割算法  

Normalized cut segmentation algorithm combined with wavelet coefficient

作　　者：黄颖[1,2] 杨光琼[1] 

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065 [2]电子科技大学电子工程学院,成都610054

出　　处：《计算机应用》2011年第1期182-183,242,共3页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60842003);教育部科学技术研究重点项目(210183);重庆市教委科学技术研究项目(KJ100525)

摘　　要：为了提高normalized cut分割算法的准确率,结合小波系数调整图像的亲和矩阵。首先采用小波系数计算图像的边缘信息,然后使用边缘信息对原始图像构造一个图,求取图的拉普拉斯矩阵的前K个特征值,并对第二个特征值对应的特征向量进行分类,得到最终的分割结果。使用一些自然图像进行了验证,结果证明新方法能提取目标更为细致的细节,保留更多有用的信息。Wavelet coefficient was used to adjust the affinity matrix of the image and to get more accurate resuhs. Firstly, the edge information of the image was calculated with wavelet coefficient. Secondly, a graph was constructed for the original image with edge information, and the Laplacian matrix was obtained and the first K eigenvalues was computed, The eigenvector corresponding to the second eigenvalue was classified to get the final segmentation results. The experimental results show that the proposed method can extract more fine details, and preserve more useful information.

关 键 词：图像分割 normalized cut方法 小波系数 图分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]